diff --git a/python-package/xgboost/core.py b/python-package/xgboost/core.py
index 856da41c156d..e27876ef9169 100644
--- a/python-package/xgboost/core.py
+++ b/python-package/xgboost/core.py
@@ -246,12 +246,13 @@ def _has_cuda_array_interface(data):
 def _maybe_pandas_data(data, feature_names, feature_types,
                        meta=None, meta_type=None):
     """Extract internal data from pd.DataFrame for DMatrix data"""
-
     if not (PANDAS_INSTALLED and isinstance(data, DataFrame)):
         return data, feature_names, feature_types
+    from pandas.api.types import is_sparse
 
     data_dtypes = data.dtypes
-    if not all(dtype.name in PANDAS_DTYPE_MAPPER for dtype in data_dtypes):
+    if not all(dtype.name in PANDAS_DTYPE_MAPPER or is_sparse(dtype)
+               for dtype in data_dtypes):
         bad_fields = [
             str(data.columns[i]) for i, dtype in enumerate(data_dtypes)
             if dtype.name not in PANDAS_DTYPE_MAPPER
@@ -272,9 +273,12 @@ def _maybe_pandas_data(data, feature_names, feature_types,
             feature_names = data.columns.format()
 
     if feature_types is None and meta is None:
-        feature_types = [
-            PANDAS_DTYPE_MAPPER[dtype.name] for dtype in data_dtypes
-        ]
+        feature_types = []
+        for dtype in data_dtypes:
+            if is_sparse(dtype):
+                feature_types.append(PANDAS_DTYPE_MAPPER[dtype.subtype.name])
+            else:
+                feature_types.append(PANDAS_DTYPE_MAPPER[dtype.name])
 
     if meta and len(data.columns) > 1:
         raise ValueError(
diff --git a/tests/python/test_with_pandas.py b/tests/python/test_with_pandas.py
index 4da606269e1c..4d20ee72ca42 100644
--- a/tests/python/test_with_pandas.py
+++ b/tests/python/test_with_pandas.py
@@ -109,6 +109,22 @@ def test_pandas(self):
         assert dm.num_row() == 2
         assert dm.num_col() == 6
 
+    def test_pandas_sparse(self):
+        import pandas as pd
+        rows = 100
+        X = pd.DataFrame(
+            {"A": pd.SparseArray(np.random.randint(0, 10, size=rows)),
+             "B": pd.SparseArray(np.random.randn(rows)),
+             "C": pd.SparseArray(np.random.permutation(
+                 [True, False] * (rows // 2)))}
+        )
+        y = pd.Series(pd.SparseArray(np.random.randn(rows)))
+        dtrain = xgb.DMatrix(X, y)
+        booster = xgb.train({}, dtrain, num_boost_round=4)
+        predt_sparse = booster.predict(xgb.DMatrix(X))
+        predt_dense = booster.predict(xgb.DMatrix(X.sparse.to_dense()))
+        np.testing.assert_allclose(predt_sparse, predt_dense)
+
     def test_pandas_label(self):
         # label must be a single column
         df = pd.DataFrame({'A': ['X', 'Y', 'Z'], 'B': [1, 2, 3]})